Research map / Image generation / Text-to-image diffusion
TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering
Hanshen Zhu, Yuliang Liu, Xuecheng Wu, An-Lan Wang, Hao Feng, Dingkang Yang, Chao Feng, Can Huang, Jingqun Tang, Xiang Bai
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add TextPecker structure-aware reward in RL | Flux.1[dev] · GenTextEval English | Sem.: 0.336 -> 0.719 | Improved |
| use TextPecker reward instead of OCR reward | Flux.1[dev] · GenTextEval English | Sem.: 0.602 -> 0.719 | Improved |
| use TextPecker reward instead of OCR reward | SD3.5-M · CVTG-2K English | Sem.: 0.671 -> 0.593 | Hurt |
| use TextPecker reward instead of OCR reward on Chinese | Qwen-Image · GenTextEval Chinese | Sem.: 0.874 -> 0.897 | Improved |
| add TextPecker reward in RL | Qwen-Image · OneIG English | Avg.: 0.871 -> 0.990 | Improved |
| add synthesized data to annotated data (ablation) | TextPecker (InternVL3-8B) · TSAP Chinese | F1: 0.727 -> 0.927 | Improved |
| add synthesized data to annotated data (ablation) | TextPecker (InternVL3-8B) · CTR Chinese | recall: 0.849 -> 0.962 | Improved |
| train on annotated structural data versus base model (ablation) | InternVL3-8B · TSAP English | F1: 0.183 -> 0.874 | Improved |
| add pairwise matching to OCR reward (ablation) | Flux.1[dev] · GenTextEval-EN | Sem.: 0.602 -> 0.644 | Improved |
| add pairwise matching to OCR reward (ablation) | Flux.1[dev] · GenTextEval-EN | Qua.: 0.976 -> 0.976 | No clear effect |
| replace OCR model with TextPecker in reward (ablation) | Flux.1[dev] · GenTextEval-EN | Sem.: 0.644 -> 0.702 | Improved |
| add structural quality (SQ) reward (ablation) | Flux.1[dev] · GenTextEval-EN | Sem.: 0.702 -> 0.719 | Improved |
Closest papers by meaning
- DreamText: High Fidelity Scene Text Synthesis (CVPR 2025 · similarity 0.86)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.86)
- AnyText: Multilingual Visual Text Generation and Editing (ICLR 2024 · similarity 0.86)
- Type-R: Automatically Retouching Typos for Text-to-Image Generation (CVPR 2025 · similarity 0.86)
- ASCIIEval: Benchmarking Models' Visual Perception in Text Strings via ASCII Art (ICLR 2026 · similarity 0.85)
- Rich Human Feedback for Text-to-Image Generation (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
